skills/google/skills/gke-security/Gen Agent Trust Hub

gke-security

Pass

Audited by Gen Agent Trust Hub on Jun 25, 2026

Risk Level: SAFE
Full Analysis
  • Security Hardening Defaults: The skill promotes a 'golden path' security posture by default, enforcing features like Workload Identity Federation, Secret Manager integration, and Shielded GKE Nodes.
  • Least Privilege Access: Instructions emphasize granting specific IAM roles (e.g., roles/container.viewer) and namespace-scoped RBAC roles rather than broad administrative permissions.
  • Secure Image Usage: Verification steps and assets utilize trusted, well-known images from official registries, including gcr.io/distroless/static and gcr.io/google.com/cloudsdktool/cloud-sdk, which reduces the attack surface compared to general-purpose base images.
  • Network Security Controls: Includes templates for default-deny Network Policies and guidance for Dataplane V2, ensuring network traffic is restricted by default.
  • Workload Isolation: Provides guidance on enabling GKE Sandbox (gVisor) for running untrusted workloads in an isolated environment.
  • Indirect Prompt Injection Surface: The skill processes Kubernetes resource data and cluster configurations via tools like get_k8s_resource and get_cluster. While this represents a standard operational surface, the skill encourages verification and auditing practices to maintain a secure trust chain.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 25, 2026, 01:04 PM
Security Audit — agent-trust-hub — gke-security